廣電基金針對2006年北高市長媒體所作之民調,發現民調數據與實際開票後的比率差異過大,民調結果可能因民調機構的本身政治立場而使結果偏差,民調已是現今媒體常用的工具之一,但其可信度卻為人所詬病。 然而在這網路發達時代,各部落格文章也快速增加,人們也勇於發表自己的言論,如能將部落格文章收集並分類,也可達到自動統計民調之效,可做為第三方民調的參考。藉由現今的資訊技術支援,使得文件得以有效分類。本研究將利用資料探勘技術中的分類技術,建立適合分類器,將部落格文章分類,以做為民調結果的另一參考. 本研究用無名小站部落格文章做為資料來源,將文章資料人工判別類別做為分類評估,並藉由其文章資料收集分析關鍵詞彙,接著利用分類技術貝氏網路建立分類器。透過本研究的結果後,期望達到更快速且更準確的民調參考依據。
The Broadcasting Development Fund found that for the 2006 Taipei and Kaohsiung mayor elections, the results of the media polls were significantly different from the actual election results. The political position of the organization conducting the polls may have affected the results, polls are now one of the tools most used by the media, but its reliability is now questioned. However, in this era of advanced Internet, the number of different blog articles is quickly increasing, and writers are actively expressing their views. If these articles can be collected and classified, it can also automatically have the effect of an opinion poll, and can be used as a third-party poll reference. Through the support of modern information technology, it can allow effective document classification. This study will use the classification technology of data mining technology to establish an appropriate classification device. This device will be used to classify blog articles, and the results can serve as another opinion poll reference. This study uses Wretch Blogs as the information source; using manual labor to evaluate and categorize the article information. The keywords will then be analyzed from the collected article information, and a classifier will be established through classification Bayesian network technology. Through the results of this study, we expect to achieve more rapid and more accurate poll references.